clade-embeddings-search
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill contains educational content and code snippets for using official Anthropic SDKs. No malicious patterns, persistence mechanisms, or unauthorized privilege escalations were found.
- [INDIRECT_PROMPT_INJECTION]: The skill describes an architecture for agentic loops that process external data. 1. Ingestion points: Data from external tools (e.g., weather API) is returned to the agent in SKILL.md. 2. Boundary markers: Absent in instructional snippets. 3. Capability inventory: The skill itself is informational; actual tool capabilities are developer-defined. 4. Sanitization: Absent beyond standard JSON serialization.
Audit Metadata